European Institute, London School of Economics
June 2025
Voting → Attitudes
Growing evidence that voting affects citizens:
But: Evidence on desirable democratic outcomes is mixed — largely null effects on political knowledge, efficacy, and civic engagement (Holbein and Rangel 2020)
Taken together, these findings might suggest that voting’s primary downstream effects may not be on positive civic virtues, but rather on citizens’ emotional relationships with political groups, when we talk about attitudes, and habits when we talk about behaviour.
Under a group theory perspective, voting is a declaration of allegiance to a political group which shapes how citizens view themselves and their opponents.
Evidence:
Importantly, one can experience cognitive dissonance even when their behaviour does not fundamentally deviate from their attitudes.
Why first-time voters are crucial:
Cognitive dissonance in early voting experiences takes the form of cognitive coherence
For first time voters:
The cost of changing their opinion is very low because they have no priors
Making voting especially consequential → easier to stick because it fills this attitudinal void
US vs THEM?
US vs THEM?
Voting ? Attitudes
CSES Wave 5 (2016-2021):
Cross-national comparison of European democracries
Running Variable: Year of birth
Treatment: Voting in previous salient election
Outcome: Affective Polarisation (Wagner 2021) and Partisanship
Estimation: 2SLS (optimal bandwidth) and robust bias-corrected inference (Calonico, Cattaneo and Titiunik, 2014)
UKHLS:
Multiple Elections (2010, and 2019) and Party system transformation (Wave 2,11,and 12)
Running Variable: Month and year of birth
Treatment: Voting in current election
Outcome: Affective Polarisation (3 vs. 6 parties)
Estimation: 2SLS (optimal bandwidth) and robust bias-corrected inference
Main Finding: Voting in salient elections significantly increases affective polarization (both unweighted and weighted)
Effect size: 0.19-point increase in affective polarization, significant at all significance levels across bandwidth specifications > +/- 2 years
| ITT Effect | 2SLS Optimal | |
|---|---|---|
| (Intercept) | 0.95* | |
| [0.92; 0.98] | ||
| voting_eligibility | -0.02 | |
| [-0.08; 0.03] | ||
| year_cen | -0.00 | |
| [-0.01; 0.01] | ||
| salient_voting | -0.02 | |
| [-0.07; 0.02] | ||
| Num. obs. | 6913 | 5402 |
Wave 2 (2010): Two-Party Dominance Era
| RD (Sharp) | ITT Effect | RD (Fuzzy) | 2SLS | |
|---|---|---|---|---|
| Eligibility | 0.477** (0.160) | 0.312* (0.128) | – | – |
| Voting | – | – | 0.653** (0.222) | 0.402** (0.147) |
| N | 1393 | 1393 | 1983 | 1983 |
| Bandwidth | 5.71 | 5.71 | 9.10 | 9.10 |
| RD (Sharp) | ITT Effect | RD (Fuzzy) | 2SLS | |
|---|---|---|---|---|
| Eligibility | 0.501** (0.171) | 0.339* (0.136) | – | – |
| Voting | – | – | 0.679** (0.238) | 0.421** (0.157) |
| N | 1391 | 1391 | 1960 | 1960 |
| Bandwidth | 5.79 | 5.79 | 9.10 | 9.10 |
| RD (Sharp) | ITT Effect | RD (Fuzzy) | 2SLS | |
|---|---|---|---|---|
| Eligibility | 0.342* (0.147) | 0.178 (0.114) | – | – |
| Voting | – | – | 0.457* (0.208) | 0.299* (0.136) |
| N | 1469 | 1469 | 2034 | 2034 |
| Bandwidth | 5.99 | 5.99 | 9.09 | 9.09 |
| RD (Sharp) | ITT Effect | RD (Fuzzy) | 2SLS | |
|---|---|---|---|---|
| Eligibility | 0.498** (0.165) | 0.317* (0.131) | – | – |
| Voting | – | – | 0.675** (0.228) | 0.410** (0.151) |
| N | 1382 | 1382 | 1969 | 1969 |
| Bandwidth | 5.74 | 5.74 | 9.14 | 9.14 |
| RD (Sharp) | ITT Effect | RD (Fuzzy) | 2SLS | |
|---|---|---|---|---|
| Eligibility | 0.164 (0.119) | 0.245** (0.095) | – | – |
| Voting | – | – | 0.287 (0.193) | 0.556*** (0.142) |
| N | 2578 | 2578 | 2888 | 2888 |
| Bandwidth | 6.19 | 6.19 | 7.17 | 7.17 |
| RD (Sharp) | ITT Effect | RD (Fuzzy) | 2SLS | |
|---|---|---|---|---|
| Eligibility | 0.168 (0.128) | 0.265* (0.105) | – | – |
| Voting | – | – | 0.309 (0.205) | 0.616*** (0.151) |
| N | 2443 | 2443 | 2849 | 2849 |
| Bandwidth | 5.88 | 5.88 | 7.18 | 7.18 |
| RD (Sharp) | ITT Effect | RD (Fuzzy) | 2SLS | |
|---|---|---|---|---|
| Eligibility | 0.066 (0.112) | 0.137 (0.091) | – | – |
| Voting | – | – | 0.129 (0.185) | 0.355** (0.137) |
| N | 2753 | 2753 | 3135 | 3135 |
| Bandwidth | 6.11 | 6.11 | 7.29 | 7.29 |
| RD (Sharp) | ITT Effect | RD (Fuzzy) | 2SLS | |
|---|---|---|---|---|
| Eligibility | 0.154 (0.122) | 0.254* (0.100) | – | – |
| Voting | – | – | 0.287 (0.196) | 0.589*** (0.144) |
| N | 2462 | 2462 | 2873 | 2873 |
| Bandwidth | 5.85 | 5.85 | 7.17 | 7.17 |
Figure 1: Effects of Voting Eligibility and Voting on Affective Polarization Across Waves and Specifications
Preliminary Takeaways:
Context matters: Effects conditional on party system structure — voting seems to produce polarisation in multi-party contexts but not in simpler two-party systems.
Limitations:
Refining the Estimation:
Understanding Mechanisms:
Institutional variation:
Austria (voting age 16) and Brazil (compulsory 18-70, optional 16-17) and Germany (16-17 year-olds vote in municipal but not state/federal elections)
Compulsory voting contexts: Australia, Belgium, Brazil